On convex sets in general position (Q1067612): Difference between revisions

Convex cones \(A_ 1,...,A_ n\) from a real linear space E are called to be in general position [see \textit{S. S. Kutateladze}, Sov. Math., Dokl. 18, 146-148 (1977); translation from Dokl. Akad. Nauk SSSR 232, 770-772 (1977; Zbl 0371.47055)] if there exists a rearrangement of the cones \(A_{i_ 1},...,A_{i_ n}\) such that \((A_{i_ 1}\cap...\cap A_{i_ k})-A_{i_{k+1}}\) is a subspace for every k, \(1\leq k\leq n- 1.\) In this paper the author extends the notion to convex sets and gives different characterizations for the convex sets in general position. A convex set A from E is called to be in general position if \(0\in icr(\cap_{1\leq i\leq k}A_ i-A_ k),\) \(\forall k\), \(2\leq k\leq n\), where icr M means the algebraic relative interior of M. A convex set A from E is called to be in (strong) S-general position if \(0\in cor(\cap_{1\leq i\leq k}-A_ k),\) \(\forall k\), \(2\leq k\leq n\), where cor M is the algebraic interior of M, i.e. cor M\(=icr M\) if Aff M\(=E.\) By this definition of the convex set in general position it is seen that the notion does not depend on the order of the sets and, therefore, can be more useful in convex optimization theory in general and in algebraic convex programming in special.